LLParallax
commited on
Commit
•
a3b22a5
1
Parent(s):
81454a5
Update README.md
Browse filesupdate eval results and continue training
README.md
CHANGED
@@ -15,7 +15,7 @@ model-index:
|
|
15 |
type: nethack_challenge
|
16 |
metrics:
|
17 |
- type: mean_reward
|
18 |
-
value:
|
19 |
name: mean_reward
|
20 |
verified: false
|
21 |
---
|
@@ -38,7 +38,7 @@ python -m sample_factory.huggingface.load_from_hub -r LLParallax/sample_factory_
|
|
38 |
|
39 |
To run the model after download, use the `enjoy` script corresponding to this environment:
|
40 |
```
|
41 |
-
python -m
|
42 |
```
|
43 |
|
44 |
|
@@ -49,7 +49,7 @@ See https://www.samplefactory.dev/10-huggingface/huggingface/ for more details
|
|
49 |
|
50 |
To continue training with this model, use the `train` script corresponding to this environment:
|
51 |
```
|
52 |
-
python -m
|
53 |
```
|
54 |
|
55 |
Note, you may have to adjust `--train_for_env_steps` to a suitably high number as the experiment will resume at the number of steps it concluded at.
|
|
|
15 |
type: nethack_challenge
|
16 |
metrics:
|
17 |
- type: mean_reward
|
18 |
+
value: 3245.47 +/- 2691.37
|
19 |
name: mean_reward
|
20 |
verified: false
|
21 |
---
|
|
|
38 |
|
39 |
To run the model after download, use the `enjoy` script corresponding to this environment:
|
40 |
```
|
41 |
+
python -m sf_examples.nethack.enjoy_nethack --env=nethack_challenge --train_dir=./train_dir --experiment=sample_factory_human_monk
|
42 |
```
|
43 |
|
44 |
|
|
|
49 |
|
50 |
To continue training with this model, use the `train` script corresponding to this environment:
|
51 |
```
|
52 |
+
python -m sf_examples.nethack.train_nethack --env=nethack_challenge --character=mon-hum-neu-mal --num_workers=16 --num_envs_per_worker=32 batch_size=4096 --train_dir=./train_dir --experiment=sample_factory_human_monk --restart_behavior=resume --train_for_env_steps=10000000000
|
53 |
```
|
54 |
|
55 |
Note, you may have to adjust `--train_for_env_steps` to a suitably high number as the experiment will resume at the number of steps it concluded at.
|